cpython/Python/ceval_macros.h
Tian Gao 9c14ed0618
gh-107674: Improve performance of sys.settrace (GH-117133)
* Check tracing in RESUME_CHECK

* Only change to RESUME_CHECK if not tracing
2024-05-03 19:49:24 +01:00

450 lines
16 KiB
C

// Macros and other things needed by ceval.c, and bytecodes.c
/* Computed GOTOs, or
the-optimization-commonly-but-improperly-known-as-"threaded code"
using gcc's labels-as-values extension
(http://gcc.gnu.org/onlinedocs/gcc/Labels-as-Values.html).
The traditional bytecode evaluation loop uses a "switch" statement, which
decent compilers will optimize as a single indirect branch instruction
combined with a lookup table of jump addresses. However, since the
indirect jump instruction is shared by all opcodes, the CPU will have a
hard time making the right prediction for where to jump next (actually,
it will be always wrong except in the uncommon case of a sequence of
several identical opcodes).
"Threaded code" in contrast, uses an explicit jump table and an explicit
indirect jump instruction at the end of each opcode. Since the jump
instruction is at a different address for each opcode, the CPU will make a
separate prediction for each of these instructions, which is equivalent to
predicting the second opcode of each opcode pair. These predictions have
a much better chance to turn out valid, especially in small bytecode loops.
A mispredicted branch on a modern CPU flushes the whole pipeline and
can cost several CPU cycles (depending on the pipeline depth),
and potentially many more instructions (depending on the pipeline width).
A correctly predicted branch, however, is nearly free.
At the time of this writing, the "threaded code" version is up to 15-20%
faster than the normal "switch" version, depending on the compiler and the
CPU architecture.
NOTE: care must be taken that the compiler doesn't try to "optimize" the
indirect jumps by sharing them between all opcodes. Such optimizations
can be disabled on gcc by using the -fno-gcse flag (or possibly
-fno-crossjumping).
*/
/* Use macros rather than inline functions, to make it as clear as possible
* to the C compiler that the tracing check is a simple test then branch.
* We want to be sure that the compiler knows this before it generates
* the CFG.
*/
#ifdef WITH_DTRACE
#define OR_DTRACE_LINE | (PyDTrace_LINE_ENABLED() ? 255 : 0)
#else
#define OR_DTRACE_LINE
#endif
#ifdef HAVE_COMPUTED_GOTOS
#ifndef USE_COMPUTED_GOTOS
#define USE_COMPUTED_GOTOS 1
#endif
#else
#if defined(USE_COMPUTED_GOTOS) && USE_COMPUTED_GOTOS
#error "Computed gotos are not supported on this compiler."
#endif
#undef USE_COMPUTED_GOTOS
#define USE_COMPUTED_GOTOS 0
#endif
#ifdef Py_STATS
#define INSTRUCTION_STATS(op) \
do { \
OPCODE_EXE_INC(op); \
if (_Py_stats) _Py_stats->opcode_stats[lastopcode].pair_count[op]++; \
lastopcode = op; \
} while (0)
#else
#define INSTRUCTION_STATS(op) ((void)0)
#endif
#if USE_COMPUTED_GOTOS
# define TARGET(op) TARGET_##op:
# define DISPATCH_GOTO() goto *opcode_targets[opcode]
#else
# define TARGET(op) case op: TARGET_##op:
# define DISPATCH_GOTO() goto dispatch_opcode
#endif
/* PRE_DISPATCH_GOTO() does lltrace if enabled. Normally a no-op */
#ifdef LLTRACE
#define PRE_DISPATCH_GOTO() if (lltrace >= 5) { \
lltrace_instruction(frame, stack_pointer, next_instr, opcode, oparg); }
#else
#define PRE_DISPATCH_GOTO() ((void)0)
#endif
#if LLTRACE
#define LLTRACE_RESUME_FRAME() \
do { \
lltrace = maybe_lltrace_resume_frame(frame, &entry_frame, GLOBALS()); \
if (lltrace < 0) { \
goto exit_unwind; \
} \
} while (0)
#else
#define LLTRACE_RESUME_FRAME() ((void)0)
#endif
#ifdef Py_GIL_DISABLED
#define QSBR_QUIESCENT_STATE(tstate) _Py_qsbr_quiescent_state(((_PyThreadStateImpl *)tstate)->qsbr)
#else
#define QSBR_QUIESCENT_STATE(tstate)
#endif
/* Do interpreter dispatch accounting for tracing and instrumentation */
#define DISPATCH() \
{ \
NEXTOPARG(); \
PRE_DISPATCH_GOTO(); \
DISPATCH_GOTO(); \
}
#define DISPATCH_SAME_OPARG() \
{ \
opcode = next_instr->op.code; \
PRE_DISPATCH_GOTO(); \
DISPATCH_GOTO(); \
}
#define DISPATCH_INLINED(NEW_FRAME) \
do { \
assert(tstate->interp->eval_frame == NULL); \
_PyFrame_SetStackPointer(frame, stack_pointer); \
(NEW_FRAME)->previous = frame; \
frame = tstate->current_frame = (NEW_FRAME); \
CALL_STAT_INC(inlined_py_calls); \
goto start_frame; \
} while (0)
// Use this instead of 'goto error' so Tier 2 can go to a different label
#define GOTO_ERROR(LABEL) goto LABEL
#define CHECK_EVAL_BREAKER() \
_Py_CHECK_EMSCRIPTEN_SIGNALS_PERIODICALLY(); \
QSBR_QUIESCENT_STATE(tstate); \
if (_Py_atomic_load_uintptr_relaxed(&tstate->eval_breaker) & _PY_EVAL_EVENTS_MASK) { \
if (_Py_HandlePending(tstate) != 0) { \
GOTO_ERROR(error); \
} \
}
/* Tuple access macros */
#ifndef Py_DEBUG
#define GETITEM(v, i) PyTuple_GET_ITEM((v), (i))
#else
static inline PyObject *
GETITEM(PyObject *v, Py_ssize_t i) {
assert(PyTuple_Check(v));
assert(i >= 0);
assert(i < PyTuple_GET_SIZE(v));
return PyTuple_GET_ITEM(v, i);
}
#endif
/* Code access macros */
/* The integer overflow is checked by an assertion below. */
#define INSTR_OFFSET() ((int)(next_instr - _PyCode_CODE(_PyFrame_GetCode(frame))))
#define NEXTOPARG() do { \
_Py_CODEUNIT word = {.cache = FT_ATOMIC_LOAD_UINT16_RELAXED(*(uint16_t*)next_instr)}; \
opcode = word.op.code; \
oparg = word.op.arg; \
} while (0)
/* JUMPBY makes the generator identify the instruction as a jump. SKIP_OVER is
* for advancing to the next instruction, taking into account cache entries
* and skipped instructions.
*/
#define JUMPBY(x) (next_instr += (x))
#define SKIP_OVER(x) (next_instr += (x))
/* OpCode prediction macros
Some opcodes tend to come in pairs thus making it possible to
predict the second code when the first is run. For example,
COMPARE_OP is often followed by POP_JUMP_IF_FALSE or POP_JUMP_IF_TRUE.
Verifying the prediction costs a single high-speed test of a register
variable against a constant. If the pairing was good, then the
processor's own internal branch predication has a high likelihood of
success, resulting in a nearly zero-overhead transition to the
next opcode. A successful prediction saves a trip through the eval-loop
including its unpredictable switch-case branch. Combined with the
processor's internal branch prediction, a successful PREDICT has the
effect of making the two opcodes run as if they were a single new opcode
with the bodies combined.
If collecting opcode statistics, your choices are to either keep the
predictions turned-on and interpret the results as if some opcodes
had been combined or turn-off predictions so that the opcode frequency
counter updates for both opcodes.
Opcode prediction is disabled with threaded code, since the latter allows
the CPU to record separate branch prediction information for each
opcode.
*/
#define PREDICT_ID(op) PRED_##op
#define PREDICTED(op) PREDICT_ID(op):
/* Stack manipulation macros */
/* The stack can grow at most MAXINT deep, as co_nlocals and
co_stacksize are ints. */
#define STACK_LEVEL() ((int)(stack_pointer - _PyFrame_Stackbase(frame)))
#define STACK_SIZE() (_PyFrame_GetCode(frame)->co_stacksize)
#define EMPTY() (STACK_LEVEL() == 0)
#define TOP() (stack_pointer[-1])
#define SECOND() (stack_pointer[-2])
#define THIRD() (stack_pointer[-3])
#define FOURTH() (stack_pointer[-4])
#define PEEK(n) (stack_pointer[-(n)])
#define POKE(n, v) (stack_pointer[-(n)] = (v))
#define SET_TOP(v) (stack_pointer[-1] = (v))
#define SET_SECOND(v) (stack_pointer[-2] = (v))
#define BASIC_STACKADJ(n) (stack_pointer += n)
#define BASIC_PUSH(v) (*stack_pointer++ = (v))
#define BASIC_POP() (*--stack_pointer)
#ifdef Py_DEBUG
#define PUSH(v) do { \
BASIC_PUSH(v); \
assert(STACK_LEVEL() <= STACK_SIZE()); \
} while (0)
#define POP() (assert(STACK_LEVEL() > 0), BASIC_POP())
#define STACK_GROW(n) do { \
assert(n >= 0); \
BASIC_STACKADJ(n); \
assert(STACK_LEVEL() <= STACK_SIZE()); \
} while (0)
#define STACK_SHRINK(n) do { \
assert(n >= 0); \
assert(STACK_LEVEL() >= n); \
BASIC_STACKADJ(-(n)); \
} while (0)
#else
#define PUSH(v) BASIC_PUSH(v)
#define POP() BASIC_POP()
#define STACK_GROW(n) BASIC_STACKADJ(n)
#define STACK_SHRINK(n) BASIC_STACKADJ(-(n))
#endif
/* Data access macros */
#define FRAME_CO_CONSTS (_PyFrame_GetCode(frame)->co_consts)
#define FRAME_CO_NAMES (_PyFrame_GetCode(frame)->co_names)
/* Local variable macros */
#define LOCALS_ARRAY (frame->localsplus)
#define GETLOCAL(i) (frame->localsplus[i])
/* The SETLOCAL() macro must not DECREF the local variable in-place and
then store the new value; it must copy the old value to a temporary
value, then store the new value, and then DECREF the temporary value.
This is because it is possible that during the DECREF the frame is
accessed by other code (e.g. a __del__ method or gc.collect()) and the
variable would be pointing to already-freed memory. */
#define SETLOCAL(i, value) do { PyObject *tmp = GETLOCAL(i); \
GETLOCAL(i) = value; \
Py_XDECREF(tmp); } while (0)
#define GO_TO_INSTRUCTION(op) goto PREDICT_ID(op)
#ifdef Py_STATS
#define UPDATE_MISS_STATS(INSTNAME) \
do { \
STAT_INC(opcode, miss); \
STAT_INC((INSTNAME), miss); \
/* The counter is always the first cache entry: */ \
if (ADAPTIVE_COUNTER_TRIGGERS(next_instr->cache)) { \
STAT_INC((INSTNAME), deopt); \
} \
} while (0)
#else
#define UPDATE_MISS_STATS(INSTNAME) ((void)0)
#endif
#define DEOPT_IF(COND, INSTNAME) \
if ((COND)) { \
/* This is only a single jump on release builds! */ \
UPDATE_MISS_STATS((INSTNAME)); \
assert(_PyOpcode_Deopt[opcode] == (INSTNAME)); \
GO_TO_INSTRUCTION(INSTNAME); \
}
#define GLOBALS() frame->f_globals
#define BUILTINS() frame->f_builtins
#define LOCALS() frame->f_locals
#define CONSTS() _PyFrame_GetCode(frame)->co_consts
#define NAMES() _PyFrame_GetCode(frame)->co_names
#define DTRACE_FUNCTION_ENTRY() \
if (PyDTrace_FUNCTION_ENTRY_ENABLED()) { \
dtrace_function_entry(frame); \
}
/* This takes a uint16_t instead of a _Py_BackoffCounter,
* because it is used directly on the cache entry in generated code,
* which is always an integral type. */
#define ADAPTIVE_COUNTER_TRIGGERS(COUNTER) \
backoff_counter_triggers(forge_backoff_counter((COUNTER)))
#ifdef Py_GIL_DISABLED
#define ADVANCE_ADAPTIVE_COUNTER(COUNTER) \
do { \
/* gh-115999 tracks progress on addressing this. */ \
static_assert(0, "The specializing interpreter is not yet thread-safe"); \
} while (0);
#else
#define ADVANCE_ADAPTIVE_COUNTER(COUNTER) \
do { \
(COUNTER) = advance_backoff_counter((COUNTER)); \
} while (0);
#endif
#define PAUSE_ADAPTIVE_COUNTER(COUNTER) \
do { \
(COUNTER) = pause_backoff_counter((COUNTER)); \
} while (0);
#define UNBOUNDLOCAL_ERROR_MSG \
"cannot access local variable '%s' where it is not associated with a value"
#define UNBOUNDFREE_ERROR_MSG \
"cannot access free variable '%s' where it is not associated with a value" \
" in enclosing scope"
#define NAME_ERROR_MSG "name '%.200s' is not defined"
#define DECREF_INPUTS_AND_REUSE_FLOAT(left, right, dval, result) \
do { \
if (Py_REFCNT(left) == 1) { \
((PyFloatObject *)left)->ob_fval = (dval); \
_Py_DECREF_SPECIALIZED(right, _PyFloat_ExactDealloc);\
result = (left); \
} \
else if (Py_REFCNT(right) == 1) {\
((PyFloatObject *)right)->ob_fval = (dval); \
_Py_DECREF_NO_DEALLOC(left); \
result = (right); \
}\
else { \
result = PyFloat_FromDouble(dval); \
if ((result) == NULL) GOTO_ERROR(error); \
_Py_DECREF_NO_DEALLOC(left); \
_Py_DECREF_NO_DEALLOC(right); \
} \
} while (0)
// If a trace function sets a new f_lineno and
// *then* raises, we use the destination when searching
// for an exception handler, displaying the traceback, and so on
#define INSTRUMENTED_JUMP(src, dest, event) \
do { \
if (tstate->tracing) {\
next_instr = dest; \
} else { \
_PyFrame_SetStackPointer(frame, stack_pointer); \
next_instr = _Py_call_instrumentation_jump(tstate, event, frame, src, dest); \
stack_pointer = _PyFrame_GetStackPointer(frame); \
if (next_instr == NULL) { \
next_instr = (dest)+1; \
goto error; \
} \
} \
} while (0);
// GH-89279: Force inlining by using a macro.
#if defined(_MSC_VER) && SIZEOF_INT == 4
#define _Py_atomic_load_relaxed_int32(ATOMIC_VAL) (assert(sizeof((ATOMIC_VAL)->_value) == 4), *((volatile int*)&((ATOMIC_VAL)->_value)))
#else
#define _Py_atomic_load_relaxed_int32(ATOMIC_VAL) _Py_atomic_load_relaxed(ATOMIC_VAL)
#endif
static inline int _Py_EnterRecursivePy(PyThreadState *tstate) {
return (tstate->py_recursion_remaining-- <= 0) &&
_Py_CheckRecursiveCallPy(tstate);
}
static inline void _Py_LeaveRecursiveCallPy(PyThreadState *tstate) {
tstate->py_recursion_remaining++;
}
/* Implementation of "macros" that modify the instruction pointer,
* stack pointer, or frame pointer.
* These need to treated differently by tier 1 and 2.
* The Tier 1 version is here; Tier 2 is inlined in ceval.c. */
#define LOAD_IP(OFFSET) do { \
next_instr = frame->instr_ptr + (OFFSET); \
} while (0)
/* There's no STORE_IP(), it's inlined by the code generator. */
#define LOAD_SP() \
stack_pointer = _PyFrame_GetStackPointer(frame);
/* Tier-switching macros. */
#ifdef _Py_JIT
#define GOTO_TIER_TWO(EXECUTOR) \
do { \
OPT_STAT_INC(traces_executed); \
jit_func jitted = (EXECUTOR)->jit_code; \
next_instr = jitted(frame, stack_pointer, tstate); \
Py_DECREF(tstate->previous_executor); \
tstate->previous_executor = NULL; \
frame = tstate->current_frame; \
if (next_instr == NULL) { \
goto resume_with_error; \
} \
stack_pointer = _PyFrame_GetStackPointer(frame); \
DISPATCH(); \
} while (0)
#else
#define GOTO_TIER_TWO(EXECUTOR) \
do { \
OPT_STAT_INC(traces_executed); \
next_uop = (EXECUTOR)->trace; \
assert(next_uop->opcode == _START_EXECUTOR || next_uop->opcode == _COLD_EXIT); \
goto enter_tier_two; \
} while (0)
#endif
#define GOTO_TIER_ONE(TARGET) \
do { \
Py_DECREF(tstate->previous_executor); \
tstate->previous_executor = NULL; \
next_instr = target; \
DISPATCH(); \
} while (0)
#define CURRENT_OPARG() (next_uop[-1].oparg)
#define CURRENT_OPERAND() (next_uop[-1].operand)
#define JUMP_TO_JUMP_TARGET() goto jump_to_jump_target
#define JUMP_TO_ERROR() goto jump_to_error_target
#define GOTO_UNWIND() goto error_tier_two
#define EXIT_TO_TRACE() goto exit_to_trace
#define EXIT_TO_TIER1() goto exit_to_tier1
#define EXIT_TO_TIER1_DYNAMIC() goto exit_to_tier1_dynamic;